You should the type and class of meter you're using as well as specifically how you're using your SPL meter. There are multiple types of SPL meters as well as classes which offer different information and use different ways to measure which are weighted differently. Both the audio material you're attempting to measure and the method of measurement will play heavy roles in the result. Additionally, if you're using music to measure your SPL, this will return different values based on whether you're measuring based on peaks or averages. Peak SPL will typically have a 6dB crest above the average where your transient peaks are in music (kick drum/crash cymbal/etc.) as they demand more power/SPL to be heard as a percussive sound over the average level. Conversational speech has average levels at roughly 60dB SPL, so background music would be considered around this level on average, placing peaks at about 65dB. It seems you were likely measuring an A-weighted average SPL of music, which will discriminate/ignore much of the low frequency, and return a slightly deflated average level as it's only measuring from about 500Hz to 10kHz. If you read 65dB, I would estimate you're actually hitting somewhere in the low to mid 70's for peak SPL, and an average level in the upper 60's. If you want to get an accurate reading of what SPL is in terms of "loudness", I would recommend using audio material with no transients and no bandwidth discrimination as well as a weighted meter which has no bandwidth discrimination and can implement peak monitoring. The best way to do this is use pink noise (equal energy per octave), and an RTA meter with either C or Z weighting. White noise is also usable (equal energy per frequency), but you'd need to find an FFT-based meter, which are difficult to find.
On a side note, phone microphones will rarely get an accurate reading because they are bandwidth-limited and don't have smooth transfer functions for accurate readings.